Diese Präsentation wurde erfolgreich gemeldet.
Wir verwenden Ihre LinkedIn Profilangaben und Informationen zu Ihren Aktivitäten, um Anzeigen zu personalisieren und Ihnen relevantere Inhalte anzuzeigen. Sie können Ihre Anzeigeneinstellungen jederzeit ändern.
Semantic Image Synthesis with Spatially-
Adaptive Normalization
Taesung Park1∗ Ming-Yu Liu2 Ting-Chun Wang2 Jun-Yan Zhu2,3...
Motivation
• The conventional network architecture, which is built by stacking
convolutional, normalization, and nonlinear...
1. Replace the segmentation mask m with the image class label and
making the modulation parameters spatially-invariant——
C...
Why does SPADE work better?
• It can better preserve semantic information against common
normalization layers
Semantic Image Synthesis with Spatially-Adaptive Normalization
Semantic Image Synthesis with Spatially-Adaptive Normalization
Semantic Image Synthesis with Spatially-Adaptive Normalization
Semantic Image Synthesis with Spatially-Adaptive Normalization
Semantic Image Synthesis with Spatially-Adaptive Normalization
Semantic Image Synthesis with Spatially-Adaptive Normalization
Semantic Image Synthesis with Spatially-Adaptive Normalization
Nächste SlideShare
Wird geladen in …5
×

Semantic Image Synthesis with Spatially-Adaptive Normalization

40 Aufrufe

Veröffentlicht am

yutian lin

Veröffentlicht in: Technologie
  • Als Erste(r) kommentieren

  • Gehören Sie zu den Ersten, denen das gefällt!

Semantic Image Synthesis with Spatially-Adaptive Normalization

  1. 1. Semantic Image Synthesis with Spatially- Adaptive Normalization Taesung Park1∗ Ming-Yu Liu2 Ting-Chun Wang2 Jun-Yan Zhu2,3 1UC Berkeley 2NVIDIA 3MIT CSAIL https://nvlabs.github.io/SPADE/
  2. 2. Motivation • The conventional network architecture, which is built by stacking convolutional, normalization, and nonlinearity layers, is at best sub- optimal, because their normalization layers tend to “wash away” information in input semantic masks.
  3. 3. 1. Replace the segmentation mask m with the image class label and making the modulation parameters spatially-invariant—— Conditional Batch Normalization Layer 2. Replace the segmentation mask with another image, making the modulation parameters spatially invariant and setting N = 1 —— AdaIN
  4. 4. Why does SPADE work better? • It can better preserve semantic information against common normalization layers

×